37 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512vl | FileCheck %s3 4declare <16 x i32> @llvm.experimental.vector.compress.v16i32(<16 x i32>, <16 x i1>, <16 x i32>)5 6define <16 x i32> @test_compress_freeze_elimination(<16 x i32> %a0, <16 x i32> %a1, <16 x i8> %a3) {7; CHECK-LABEL: test_compress_freeze_elimination:8; CHECK: # %bb.0:9; CHECK-NEXT: vpcmpgtd %zmm1, %zmm0, %k110; CHECK-NEXT: vpmovzxbd {{.*#+}} zmm1 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero11; CHECK-NEXT: vpbroadcastd {{.*#+}} zmm0 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]12; CHECK-NEXT: vpcompressd %zmm1, %zmm0 {%k1}13; CHECK-NEXT: retq14 %cmp = icmp sgt <16 x i32> %a0, %a115 %ext = zext <16 x i8> %a3 to <16 x i32>16 %cpr = call <16 x i32> @llvm.experimental.vector.compress.v16i32(<16 x i32> %ext, <16 x i1> %cmp, <16 x i32> splat(i32 15))17 %fr = freeze <16 x i32> %cpr18 %and = and <16 x i32> %fr, splat(i32 255)19 ret <16 x i32> %and20}21 22define <16 x i32> @test_compress_freeze(<16 x i32> %a0, <16 x i32> %a1, <16 x i8> %a3) {23; CHECK-LABEL: test_compress_freeze:24; CHECK: # %bb.0:25; CHECK-NEXT: vpcmpgtd %zmm1, %zmm0, %k126; CHECK-NEXT: vpmovzxbd {{.*#+}} zmm0 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero,xmm2[4],zero,zero,zero,xmm2[5],zero,zero,zero,xmm2[6],zero,zero,zero,xmm2[7],zero,zero,zero,xmm2[8],zero,zero,zero,xmm2[9],zero,zero,zero,xmm2[10],zero,zero,zero,xmm2[11],zero,zero,zero,xmm2[12],zero,zero,zero,xmm2[13],zero,zero,zero,xmm2[14],zero,zero,zero,xmm2[15],zero,zero,zero27; CHECK-NEXT: vpcompressd %zmm0, %zmm0 {%k1}28; CHECK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm029; CHECK-NEXT: retq30 %cmp = icmp sgt <16 x i32> %a0, %a131 %ext = zext <16 x i8> %a3 to <16 x i32>32 %cpr = call <16 x i32> @llvm.experimental.vector.compress.v16i32(<16 x i32> %ext, <16 x i1> %cmp, <16 x i32> poison)33 %fr = freeze <16 x i32> %cpr34 %and = and <16 x i32> %fr, splat(i32 255)35 ret <16 x i32> %and36}37